【作者】 耿超；

【导师】 刘杰；

【作者基本信息】 中国科学院研究生院（近代物理研究所） ， 粒子物理与原子核物理， 2014， 博士

【摘要】 随着国家航天事业的发展和战略需求的增加，半导体器件的抗辐射性能问题受到了广泛的研究，而单粒子效应作为影响宇航器件可靠性的关键因素已引起了越来越多关注。当今主流工艺的半导体器件均为微纳级尺寸，其对单粒子效应的敏感性也随之增加。而传统的单粒子效应抗辐射可靠性工程中的多种因素在一定程度上受到了质疑且需要进一步考察，尤其是经典模型中微纳级器件由于集成度原因带来的新问题(比如：边缘效应变得严重等)，以及粒子特性和器件特性与单粒子效应相互依赖关系。本文基于理论模拟的研究和微纳级器件的新特点，对微纳级SRAM器件单粒子效应的微观机理及其多种影响因素展开了深入而系统的研究。主要的研究内容与成果如下：(1)构建了适用于不同特征尺寸SRAM器件单粒子效应评估的多功能程序MUFPSA (Multi-Functional Package for SEEs Analysis)。基于Monte Carlo模拟的开源软件GEANT4，给出了对粒子特性、器件特性、作用过程等可变参量的单粒子效应分析工具。其主要功能特点在于结合了经典RPP模型、粒子入射器件几何单元后电荷扩散与收集模型、器件几何结构灵活性与多变性、SRAM阵列单元多样性和单粒子多位翻转的探测与图形记录等。运用MUFPSA程序可实现对航天元器件抗辐照特性部分功能的预评估以及对测试结果的分析和理论指导。(2)量化了微纳级SRAM器件中灵敏单元尺寸对单粒子翻转效应的影响与其边缘效应引起的能量沉积差异性。研究了灵敏单元尺寸与单粒子翻转效应预估模型的关系，计算并分析了粒子入射后的电荷产生、漂移和扩散等过程，探讨了由漏斗效应衍生出的‘漏斗长度’是否应当作为一个关键的修正参数引入经典模型。另外，对沿器件横纵方向上尺寸变量下的边缘效应引起的能量沉积信息差异性进行了定量研究与定性分析。研究结果表明，a）宇航半导体器件的在轨翻转率计算对器件灵敏区体积尺寸参数具有较为强烈的依赖性。采用RPP模型计算器件在轨翻转率时，应考虑灵敏体积尺寸参数P的大小。在耗尽层厚度0<h≤4μm前提下，为减少灵敏区LET值变化引起的系列误差，选取P≥1时的RPP模型能获得更加合理和准确的计算结果，而漏斗深度在仿真计算过程中应如何取舍和选用，则更大程度上取决于特定器件的结构以及掺杂浓度等因素；b）9.5MeV/u209Bi和20MeV/u132Xe入射引起的能量沉积在表面积和厚度参量变化时的规律基本一致。同时，相比较而言，能量沉积对表面积比对厚度参量更加敏感，且随着入射角度的增加，能量沉积更加依赖于表面积。相反，受有限的电荷密度和几何临界角度的限制，随着厚度参量的变化，能量沉积的差值最终趋向于饱和值。(3)研究了粒子特性与微纳级SRAM器件单粒子翻转效应的敏感性。一方面，通过对影响SEU/MBU敏感性的因素的分析（包括临界电荷大小、临近单元间距、布线层结构等），探究了倾角入射时单粒子多位翻转效应的空间特性，。另一方面，基于离子径迹径向特征的分析，对常用的抗辐射考核指标LET的适用性进行了定量研究。具体说来，分别以单一单元与阵列单元为研究对象，对相同LET值不同粒子特性的辐射环境诱发单位与多位翻转效应的异同性进行了深入分析。研究结果表明，a）随着临界电荷量和临近单元间距的减小，单粒子多位翻转事件率和翻转图形多样性均有所增加，同时入射角度的增大也可产生同样的效果；b）器件尺寸与单粒子多位翻转效应的敏感性更加依赖于测试器件的结构，尤其是布线层；c)多位翻转率在相同LET值下的异同使得单独使用LET参量作为单粒子效应考核量尚欠妥当。并且，离子径迹特征体现了径迹半径与粒子速度，二者在一定程度上反映了单粒子多位翻转发生的概率。此研究中粒子触发多位翻转的能力依次为：185.08MeV132Xe>140.63MeV209Bi>1231.33MeV132Xe>7625.26MeV209Bi，这主要源于低速粒子引起径迹芯周围电子空穴对密度相对较高，从而在单位面积内电荷收集较多。(4)研究了由质子与重离子诱发的单粒子效应，重点探讨了低能质子引起45nmSRAM器件的单粒子翻转效应。从辐射损伤机理的角度分析了质子与重离子引起单粒子效应的异同性；结合在轨翻转率模型的桥梁作用，建立了质子与重离子入射的等效性关系；应用MUFPSA对低能质子引起45nm SRAM器件单粒子翻转效应的新现象进行了阐述和剖析。研究结果表明，a)基于单粒子效应发生机制的一致性，质子与重离子引起的单粒子翻转截面表现了一定的关联性，主要体现在了低能质子直接电离的高饱和截面曲线与低能重离子核反应的低饱和截面曲线; b)基于单粒子在轨翻转率预测模型，重离子加速器试验结果可作为支撑数据用来计算质子引起单粒子效应的翻转截面；c)在45nm SRAM器件中低能质子可引起单粒子效应，且能量低于1MeV下引起的单粒子多位翻转效应与质子能量损失、能量沉积和单元尺寸特性均有关系。(5)建立了系统的单粒子效应微观机理分析方法。基于单粒子效应中存在的多种问题和普遍采用的分析手段，运用MUFPSA对能量离散特性、离子径迹径向特征、粒子入射瞬态特性及边缘效应误差分析等进行了系统地阐述，并讨论了它们在单粒子效应微观机理分析中的适用对象与范围。研究结果表明，a)能量离散特性与粒子种类、能量有着密切的关系，它能够为粒子特性在单粒子效应分析中引起的误差提供一定的数据和支撑；b)离子径迹径向特性对单粒子多位翻转效应的翻转率与翻转图形分析起着重要作用，尤其是研究对象为小尺寸器件时；c)粒子入射瞬态特性是对粒子入射到器件后时间和脉冲特性的描述，它能够对能量沉积与微观输运过程的关系进行定量与定性分析；d)边缘效应误差分析方法是以能量沉积为载体，针对微纳级器件单粒子效应中出现翻转截面异常现象的补充和支撑。更多还原

【Abstract】 With the development and strategic requirement of national aerospace industry,the radiation response of underlying integrated circuits and semiconductor devices todifferent radiation environment has been widely studied and investigated in the fieldof radiation effects and the related radiation-hardened technology project. Singleevent effects (SEEs), as a vital factor to result in parametric degradation or functionalfailure in devices exposure to radiation resources, has been concerned andinvestigated in areas of defense and space navigation. The currently usedsemiconductor devices are mainly in the technological scale of Micro/Nano, whichare more sensitive to the SEEs susceptibility than that of conventional sizes. Forinstance, edge effect within SEEs becomes more serious when the device features arescaling down to Micro/Nano, and the ion profile and device internal property hasessential influence on the SEEs.This dissertation has characterized, examined and analyzed the single event upsetsensitivity of static random access memories (SRAMs) of Micro/Nano scales and theunderlying physical mechanism, based on the theoretical model/computer simulation.The main research approaches and results of this dissertation are as follows,(1) The MUlti-Functional Package for SEEs Analysis (MUFPSA) has beensuccessfully constructed to characterize the SEEs evaluation and assessment. Wehave programmed the package to approach the performance including thevariances of ion-beam profile, device property, physical process, etc. Based on theGEANT4, it is necessary to point out that advantages of MUFPSA are mainlypresented about integration of classical model of Rectangular Paralleled-piped (RPP) and charge diffusion-collection process, flexibility for altering devicegeometry, SRAMs array construction and detection of multiple-bit upset (MBU)probability/multiplicity. With the aid of MUFPSA, the devices applied inaerospace can be evaluated and assessed in advance, and these correspondingresults als provide the supplemental evidences and theoretical guidance for SEEscharacterization.(2) Influence of the dimensions of sensitive volume (SV) on single event upset (SEU)occurrence have been identified and the edge effect aroused by it at obliqueincidence has been quantified as the difference of deposited energy between theincident way of randomly strike and traverse on the center of SV. It is proven thatthe SEU rate predictiondepends to some extent on the dimensions of SV, butconcerning the funnel effect, it is more likely rely on the specific property ofdevice. In addition, the difference of deposited energy induced by differentincident ways has been quantitatively and qualitatively investigated on thecondition that the size varied at the vertical and horizontal direction. The resultsshow that, a) the on-orbit SEU rate of aerospace semiconductor devices has thedependence on the dimensions of SV. Additionally, it is noted that the parameterof P should be taken into consideration when using the RPP model to predict theSEU rates. As the thickness of depletion layer is ranged from0to4μm, tomitigate the serial errors from the LET, the parameter of P should be selected asno less than1and hence the SEU rate prediction is more accurate and reasonable.The hypothetical length of funnel relys on the device structure and internalparameters; b) the regularity of loss of deposited energy induced by9.5MeV/u209Bi and20MeV/u132Xe respectively are almost identical for a certain surfacearea or thickness variances. However, it is observed that the difference ofdeposited energy is more sensitive to the factor of surface area of SV than to thevariance of thickness, and as increasing the incident angle, the trend of curves inthe sequence of the deposited energy relys more on the surface area. In contrary,the difference of deposited energy becomes terminally saturation in the case of changeable thickness, because of the limited electron-hole density and therestrained critical angle in geometry.(3) The relationship between SEU susceptibility in SRAMs within Micro/Nano scalesand ion-beam profile has been modeled. Primarily, we have explored the influencefactors about critical charge, distance between adjacent cells and over-layer onSEU/MBU sensitivity. Moreover, according to the radial profile of ion track, theapplication of LET as a metric for characterizing SEEs has been discussed. Usingthe same value of LET but different species and energy of incident ions todistinguish the radiation response of Micro/Nano scales SRAMs, the results havebeen presented on the single cell and multiple cells, respectively. The results showthat, a) the orientation of ion beams and device with different critical charge exertindispensable effects on MBU. Additionally, with the decrease of spacing distancebetween adjacent cells or the dimensions of the cells, the device is moresusceptible to SEEs, especially to MBU at oblique incidence; b) the influence ofdimensions of device on MBU sensitivity is more involved in device structure,especially its over-layers; c) LET alone is not accurate enough to characterizeSEU because MBU distribution induced by incident ions with the same LET aredifferent, particularly reflected in the MBU probability. Furthermore, twoparameters derived from ion track structure, i.e., track radius and ion relativevelocity, are taken into consideration, which turns out that the ability of incidentions triggering MBUs follows132Xe185.08MeV>209Bi140.63MeV>132Xe1231.33MeV>209Bi7625.26MeV. The explanation is that the lower ion-velocitywould produce higher radial density of e-/h+pairs and hence cause higher chargedensity.(4) The physical mechanisms of proton-induced and heavy ion-induced SEUoccurrence have been incorporated and compared with each other. It isinterestingly noted that the new phenomenon about MBU occurrence has beenobserved in45nm SRAMs, which is further analyzed through energy depositionprofile. From the standpoint of radiation damage, we have bridged theproton-induced and heavy ion-induced SEU occurrence. Based on the common-used on-orbit SEU rate prediction model, the relation of SEEs inducedby proton and heavy ion has been elucidated. It is validated that low-energyprotons have the ability to trigger SEU occurrence, and the new situation aboutMBU occurrence has been deep analyzed. The results show that, a) based on theconsistent of SEU induced by heavy ion and proton, it is observed that the SEUcross section presents the relevance in two saturation curves; b) with theintegration of the SEU rate prediction model, the heavy ion data can be insertedinto the model for equivalently predicting the rate induced by proton; c) the SEUcross-sections on the45nm SRAMs are compared with previous research work,which not only validated the simulation approach used herein, but also expose theexistence of saturated cross-section and the induced MBU when the incidentenergy is less than1MeV.(5) Technical methods for analysis of SEEs micro-mechanism have beensystematically proposed. These approaches are constructed upon the RPP model,including the deposited energy or the dispersion of it, radial ionization profile,traversing time within device, and the fixed model of collected charge at obliqueincidence. The results show that SEEs can be comprehensively characterized bythe combination of above deterministic methods. The employed approachesprovide a deep understanding of the micro-mechanism of SEEs.更多还原